AUDIO KEY LINKS: PLAYBACK DEVICES IMPROVEMENT IST PRESTO Preservation Technologies for European Broadcast Archives

Transcription

1 PRETO Preservation Technologies for European roacast Archives IT AUDIO KEY LINK: PLAYACK DEVICE IMPROVEMENT Authors: Daniele AIROLA, alvatore CANGIALOI an Giorgio Dimino (RAI)

2 2 PRETO IT D ummary 1. ummary Introuction Turntable automation Key Link escription an requirements ystem overview Aopte approach Components selection... 5 Deck an Drive motor... 5 Tone arms... 6 Microcontroller... 7 Parking Area ensors... 8 Arm Lift Motors Computer interface... 8 Physical layer... 8 Protocol Mechanical esign Chassis Arm housing Arm lift Position sensor Electronics Control harware Microcontroller firmware RPM spee implementation Turntable et up an Tuning urther Design Optimisation Ieas A/D conversion technology selection Key Link escription Equipment selection Objective measurements T 1k (manca TrueMatch a ) requency response Channel separation (crosstalk) THD+N Intermoulation Distortion (IMD) Linearity ubjective tests Objective Test methoology Test conitions Listening results Conclusions... 28

3 PRETO IT D Introuction The present Deliverable inclues a report on the results of implementation of the auio key links relate to the analog part of the transcription chain iniviuate an escribe in Deliverable 3.2, section 4.5. The analysis performe then reveale two areas of work: the ientification an assembly of a reference turntable to be employe in all the vinyl transcription chains an the selection of optimal analog to igital conversion technology. The two items touch ifferent optimisation aspects of the analog part of the transcription chain, as the former is relate to the improvement of the chain efficiency, by introucing automation an flexibility, an to the acquisition of know-how about key equipment that are isappearing from the market, while the latter has a irect impact on the quality of the signal that is capture by the transcription process, an that will be preserve in future.

4 4 PRETO IT D Turntable automation 3.1. Key Link escription an requirements Turntable automation has been ientifie as a requirement for the implementation of efficient massive vinyl transcription chains. The expecte avantage of turntable automation consists of the possibility of parallel management of several transcription lines by a single operator, as the transcription can be automatically starte with a single key press on the transcription control computer, without having to manually position the pickup on the recor, while the en is automatically etecte by a sensor, relieving the operator from the nee to continuously monitor the position of the pickup on the recor. The transcription chain for 78 RPM has further requirements as the recors were prouce in absence of stanar. everal ifferent parameters coul have been employe that require a proper matching in the reprouction equipment. The main implications are on the geometry of the pick up, on the reprouction spee an on the equalization applie. While errors on spee an equalization can be effectively compensate with a suitable igital post-processing, an therefore, be move off-line with respect to the main transcription chain, reaing the recor with an inappropriate pickup results in the loss of signal etails that cannot be recovere at a later stage. To avoi the nee for switching stylus or cartrige between the transcription of an item an the next, it is convenient to fit two tone arms on a single turntable, so that two styli configurations can be trie on the fly without any reconitioning of the chain. These requirements have been consiere critical for commercially available equipment if combine with those relative to the reprouction quality, the robustness of the equipment an the prouct an the forecast of spare parts availability in the next years. A market survey has shown that the choice of commercial turntables is rastically reuce to a scant number of semi-professional proucts if start/stop automation is require. urthermore, if two tone arms are require, the list is reuce practically to zero. ew small companies have given their availability to moify their stanar equipment accoring to user specifications, but no guarantee can be given about the continuity in time of the supply ystem overview The above requirements, fully escribe in Deliverable 3.2, bring to the ientification of a reference turntable, whose schematic representation is shown in figure 1. The platter roation is controlle by a microcontroller that sens start an stop commans to the motor an monitors the actual rotation spee. Two tonearms can be installe on the chassis, each of which equippe with an automatic arm lift an a position sensor. oth the lift motor an the position sensor are connecte to the microcontroller, which communicates with the transcription station via an R232 interface. The commans, issue by the transcription station, are interprete by the microcontroller that rives the motors that lift an lower the tonearms an controls the status of plate motor. When a tonearm reaches its parking position (en of the recor) the microcontroller etects the event by monitoring the position sensor positione on the tonearm an informs the transcription station, which in turn sens to the microcontroller the comman relative to the arm lift an stops the acquisition, provie that no auio moulation is still present on the relative pick up, so that false en of recor etections, typical of mechanical automatic stop evices, can be filtere out.

5 PRETO IT D4.3 5 Position sensor Arm lift Tonearm an pick up Auio output Position sensor Arm lift Tonearm an pick up Auio output Host interface Microcontroller an riving logic Turntable eck. Platter an rive motor igure 1. Reference turntable block iagram Aopte approach Given the ifficulty to guarantee the continuity of supply of commercial turntables compliant to the ientifie user requirements, an consiering that the total number of pieces that the archive community coul require of such equipment is probably not so high to justify the commercialisation of new proucts, it has been recognize that there is a convenience to acquire the necessary know how to assemble custom turntables, so that the assembly of small series can be commissione to local companies, on the base of a reference prototype. Therefore, the aopte approach has been that of esigning a prototype turntable, following the user requirements given in Deliverable 3.2, that can be use as a reference implementation for those who nee to implement massive vinyl transcription chains. Of course, the esign has not been starte from scratch but, rather, it has been base on the assembly of high quality parts available on the market, whose prouction can be consiere stable for the next years. pecific criteria followe in the selection of each key part are escribe in etail in the following section Components selection Deck an Drive motor Turntables can be roughly classifie accoring to two characteristics: rigi or suspene frame an belt or irect rive. roacast equipment are generally of the type with rigi frame an irect rive, see as an example EMT 960 an Technics 1200 series. The main reasons for this configuration are the higher urability of the apparatus an the cost. More in etail, rigi frames offer an intrinsic robustness against acciental shoves an are therefore more suite to be place in working environments than suspene frames that can easily be amage. On the other han, rigi frames provie poor environmental vibration insulation an must be carefully positione in orer to avoi that small knocks on the turntable boy or the supporting furniture cause vibrations that will be capture by the pick up. Direct rive motors offer fast spee stabilisation at start up an lower maintenance requirements over belt rive motors. With quartz controlle motors, the wow an flutter is extremely low an thus oesn't represent a problem, while slight more concern is given to the microfluctuation of spee ue to the PLL feeback correction. These fluctuations are better ampe with belt transmission,

6 6 PRETO IT D4.3 by means of elastic belts an high mass platters, than with irect transmission ue to the lack of an elastic element to ecouple the motor from the platter. Toay, the only professional turntable prouce in large quantities is the Technics L-1210 MK2, current revise version of the basic 1200 moel, wiely employe in the broacast an entertainment environments. The main features of this apparatus are liste in the table 1 below. Note that it oesn't fulfil our requirements uner several aspects: 78 RMP spee is not available, ouble arm configuration is not possible as there is not enough space on the eck, the position of the arm is such that recors larger than 12'' cannot be fit an there is no arm lift automation. urthermore, the external vibrations rejection has proven to be rather poor, ue to the cheap plastic frame shell an insufficient amping feet. Nevertheless, the market leaer position of Technics an the favourable quality to cost ratio make the 1210 a goo starting point for the assembly of a compliant apparatus. everal moifications nee to be applie, both to the frame an to the electronics: The original frame shell cannot be use both for the insufficient amping an because of the ifficulty to house two arms. The stanar arm has also to be remove ue to the ifficulty of installing a lift automation an in favour of a higher quality an versatile arm, as escribe in the following section. A proper frequency oscillator must be ae to implement 78 RPM spee an all the neee signal for start/stop automation must be erive. Tone arms ME eries 300 tonearms have been chosen ( This series consists of three moels, all with magnesium arm tube, iffering only for the length of the arm, erive from the more expensive V eries, from which it iffers mainly for the aition of a etachable heashell. The reasons for this choice are manifols: ME has been a market leaer in proucing high en tonearms for about 40 years an its proucts are still in prouction an easily available. The pivoting mechanism is base on ball bearings that ensures more robust an economical arms than unipivot equivalents. erie 300 has a etachable heashell that simplifies the cartrige replacement. The weight of the arms is such as to allow the use of both moving magnet (MM) an moving coil (MC) cartriges. A complete set of regulations make these arms suitable for the use with cartriges of ifferent geometries without impacting on the eck mounting. pecifically, Moel 309 (the stanar 9'' arm) will be installe in replacement of the original Technics 1210 arm, to be use for 33 an 45 RMP recor transcription, while Moel 312 (12'' long) will be mounte externally to the stanar Technics eck in the 78 RPM version, where 2 tonearms are neee an a larger playing area must be provie to play oler oversize recors. oth MC an MM cartriges can be mounte, but the high cost of MC cartriges, combine with the impossibility to substitute the stylus only, suggests that MM cartriges are a better choice for high transcription volumes. Among the choice of MM cartriges available on the market, here is a selection of moels that are suite both to the application an the aopte tonearms: Ortophon OM30 (33-45 an 78 RPM) Auio-technica AT440ML (33-45 RPM) tanton eries 500 (Custom 78 RPM tyli) hure V15VxMR (33-45 RPM)

7 PRETO IT D4.3 7 Microcontroller To relieve the transcription station from the nee to be aware of an to control all the etails of the turntable, a microcontroller installe into the turntable chassis has been use. In this way the transcription station can issue to the turntable high level commans an poll the status of the evice via a stanar serial interface, accoring to the protocol escribe in the following sections. The commans are filtere an interprete by the microcontroller that is the sole responsible for the proper functioning of the turntable, that is, the coorination between platter rotation an arms position. After a market survey, the PIC1687x series of microcontrollers from Microchip ( has been selecte because these chips have on boar all the neee peripherals for our application an the presence of an on-chip programmable flash EEPROM make the upating of the firmware fast an hany. Evaluation boars can be purchase from various venors at very low cost that can be use as a base for the turntable control with the aition of very few glue logic. In particular a PIC16877 microchip has been use, an the firmware has been evelope using the ICD-87XP emulator from Avance Transata Corporation ( The relative evaluation boar ICD-Demo has then been use for hosting the final implementation. The following table, taken from the PIC1687x ata sheet, gives the main features of the evice. Table 1. PIC1687x microcontroller features. Microcontroller Core eatures: High performance RIC CPU Only 35 single wor instructions to learn All single cycle instructions except for program branches which are two cycle Operating spee: DC - 20 M clock input DC ns instruction cycle Up to 8K x 14 wors of LAH Program Memory, Up to 368 x 8 bytes of Data Memory (RAM) Up to 256 x 8 bytes of EEPROM Data Memory Pinout compatible to the PIC16C73/74/76/77 Interrupt capability (up to 14 sources) Eight level eep harware stack Direct, inirect an relative aressing moes Power-on Reset (POR) Power-up Timer (PWRT) an Oscillator tart-up Timer (OT) Watchog Timer (WDT) with its own on-chip RC oscillator for reliable operation Programmable coe protection Power saving LEEP moe electable oscillator options Low power, high spee CMO LAH/EEPROM technology ully static esign In-Circuit erial Programming(ICP)via two pins ingle 5V In-Circuit erial Programming capability In-Circuit Debugging via two pins Processor rea/write access to program memory Wie operating voltage range: 2.0V to 5.5V High ink/ource Current: 25 ma Commercial, Inustrial an Extene temperature ranges Low-power consumption: - < 0.6 ma 3V, 4 M -20 µa 3V, 32 k - < 1 µa typical stanby current Peripheral eatures: Timer0: 8-bit timer/counter with 8-bit prescaler Timer1: 16-bit timer/counter with prescaler, can be incremente uring LEEP via external crystal/clock Timer2: 8-bit timer/counter with 8-bit perio register, prescaler an postscaler Two Capture, Compare, PWM moules - Capture is 16-bit, max. resolution is 12.5 ns -

8 8 PRETO IT D4.3 Compare is 16-bit, max. resolution is 200 ns - PWM max. resolution is 10-bit 10-bit multi-channel Analog-to-Digital converter ynchronous erial Port (P) with PI. (Master moe) an I 2 C.(Master/lave) Universal ynchronous Asynchronous Receiver Transmitter (UART/CI) with 9-bit aress etection Parallel lave Port (PP) 8-bits wie, with external RD, WR an C controls (40/44-pin only) rown-out etection circuitry for rown-out Reset (OR) Parking Area ensors The etection of en of play events, that is when the arm reaches its parking position, will be implemente using optical barriers, in orer to avoi mechanical contacts that coul be cause of problems ue to friction an wear out. Honeywell HOA2001 evices have been selecte ue to the small size an the intrinsic protection from external light erive from the use of infrare emitting ioe an etector. More information can be obtaine from the following URL: Arm Lift Motors Arm lift automation is generally not available on high en toay turntables, nor reay to use retro-fits can be foun on the market. The only viable way, therefore, is to esign a custom arm lift using commercial motors. Our market search has ientifie no motor expressly esigne for this or similar applications. Then, a step motor esigne for inustrial automation has been selecte, whose characteristics an size are compatible with the task require. Namely, we have employe a onceboz 7214 linear actuator ( evelope for the minimum regulation in a vehicle engine. This evice, being employe in consumer applications, is available in stock in quantities an offers high reliability an long lifecycle. A prototype boar with all the riving logic is also available Computer interface In the absence of a stanar proceure, a custom protocol has been efine to interface the automate turntable to the transcription control station. The main guielines followe are that of privileging the simplicity of implementation an the robustness of the communication over more evolute solutions that coul minimise an alreay low overhea at the price of higher complexity an that coul introuce critical ebugging conitions. Physical layer The communication between the computer an the turntables is one via the stanar R232 interface. Only TX an RX pins nee to be connecte, as no harware flow control signal is use by the protocol. The R232 settings are: 8 bit ata, 1 bit start, 1 bit stop, no parity, 9600 bau. Protocol The protocol is of type master-slave where the control computer acts as a master an the turntables are slaves. Each transaction is forme by a request-replay pair.

9 PRETO IT D4.3 9 The communication is always initiate by the computer by sening a byte containing one of the vali commans liste below. The replay is always constitute by a single byte. The replay will be issue "as soon as possible", typically within few ms, while the actual execution of the operations requeste coul take more time. It is responsibility of the master to enquire the evice about the completion of a comman by polling the status byte flags. Comman format A comman is constitute by a single byte issue by the master with the following format: start0 Hex 1 request to start the motor an lower arm 0 stop0 Hex 2 request to lift arm 0 an stop the motor start1 Hex 3 request to start the motor an lower arm 1 stop1 Hex 4 request to lift arm 1 an stop the motor status Hex 5 evice status request speel Hex 6 turntable spee low byte request speeh Hex 7 turntable spee high byte request all other coes are reserve Reply format A master comman must be followe by a 1 byte reply of the calle evice. The reply is one of the following: ACK NACK TATU PEEDL PEEDH Issue to acknowlege a start or stop comman Issue to reject a non vali comman or a comman issue to a non installe arm Issue in response to a status request comman Issue in response to a speel request comman Issue in response to a speeh request comman The syntax is as follows: ACK Hex 0 NACK Hex 1 TATU M L res 3 M L0 E0 L1 E1 M: 1 bit signalling if the motor is in motion (1) or stoppe (0) L0: 1 bit signalling if arm 0 is lifte (0) or on the plate (1) E0: 1 bit signalling if arm 0 has reache the parking area (1) or not (0) L1: 1 bit signalling if arm 1 is lifte (0) or on the plate (1) E1: 1 bit signalling if arm 1 has reache the parking area (1) or not (0) the remaining bits are reserve an must be set to 0 PEED

10 10 PRETO IT D4.3 The turntable spee is expresse on 2 bytes. The computer must concatenate the receive 2 bytes PEEDL an PEEDH so that a 16 bit int is forme (PEEDH is the high byte, PEEDL the low byte). The resulting short int gives the platter spee accoring to the following formula: spee (RPM) = C/(1.5*CALE*reply) where C= an CALE=2 If both the arms are use, the actual action performe by the evice to a start/stop request epens on the status of both the arms. That is, if a stop is calle for one arm while the other is still playing, the arm is lifte but the motor is not stoppe until a stop is calle also for the other arm. The evice reply will be ACK. The situation will be flagge in the TATU byte. Conversely, the motor is starte at the first start request Mechanical esign Chassis The original chassis of the Technics L-1210 turntable o not fit the requirements of this project with several respects: its rigiity is not sufficient to amp external vibrations an there is no space to house two tonearms. To avoi the complexity of reesigning the whole chassis, that requires precise housing for the platter, the electronics an the motor, all integrate in a irect rive implementation, we have opte for maintaining the original eck, mae of steel, an for substituting the plastic chassis bottom with a custom frame. The frame has been realise from a single aluminium block with an automatic milling machine. In this way, thanks to the amorphous nature of aluminium an the absence of joints, we can have a very stury an rigi frame. The table eck is then bolte to the frame. A layer of putty can be interpose to amp resiual vibrations. The isolation of the turntable from the environment is obtaine by putting the frame on rubber aseismic feet. These feet also provie regulation for the levelling of the eck. As we want to be able to play oversize recors (up to 16'' are known to have been prouce), the tonearms must be installe in such a way to provie a sufficient clear area aroun the platter. We have therefore chosen to implement the arms housings on separate elements suspene on the sies of the aluminium frame. In this way, they can be kept as options an installe only if neee, while for use with the stanar 33/45 RPM recors a ifferent housing has been esigne that replaces the original Technics arm on the eck. The etails of the implementation can be erive from igure 2 below.

11 PRETO IT D igure 2. ront view of the frame bottom. Arm housing As iscusse in the previous paragraph, two housings nee to be provie, one for the internal arm an one for the external ones. They have been both realise in mille aluminium. In the esign, special attention must be rawn to the arm position tolerances, especially concerning the arm base sieway level, that, in orer to fit the prescribe alignment between the arm an the platter, resulte to be lower than the eck plane. In any case the two planes must be maintaine parallel. The external arm housing is mae of two parts, a frame, that must be bolte to the main turntable frame, an a top plate, screwe to the frame, that houses the arm an the lift. Again, putty shoul be lai between the contact surfaces. The etails are given in the photos of igure 3 an igure 3. igure 3. Internal arm housing.

12 12 PRETO IT D4.3 igure 4. External arm housing. Arm lift The arm lift is base on the onceboz 7214 linear actuator that is provie of an antirotational evice that avois the rotation of the shaft. The motor flange is bolte uner the arm housing plate, rille so that the shaft is presente right uner the arm. A ro, screwe irectly on the motor shaft, lifts an lowers the arm, as shown in igure 5. As the shaft, while moving along its axis, tens to vibrate freely an coul shift the arm from its vertical position, a guie has been ae to the ro, that forbis any horizontal move. The provie shaft travel, 18 mm, is excessive for the application. As the motor can exert a consierable force (35 N), to avoi possible amage to the arm or the motor in case of wrong operation, two position sensors have been ae to the lower en of the shaft, so that the effective travel can be trimme accoring to the arm position regulation. The sensors have been implemente using the same type of optical barriers as the position sensor. ome harware circuitry has also been ae to electrically inhibit any shaft motion outsie the efine area. When the shaft reaches its top position, a ree glue on it breaks one of the barriers, causing an immeiate stop of the motor; in the same way, when the shaft reaches its lower position, the breaking of the other barrier stops the motor. The travel can be regulate by moifying the ree shape or sliing it along the shaft (see igure 6). igure 5. The arm lift.

13 PRETO IT D igure 6. Detail of the arm lift motor an position sensors. Position sensor The position sensor is responsible for etecting the en of recor playback. It has been implemente with an infrare switch triggere by a ree glue to the rear of the arm. The sensor is fixe on the arm pillar via a collar. Rotating the collar, it is possible to trim the etection position (see igure 7). igure 7. Parking area position sensor Electronics Control harware The control harware inclues the microcontroller, the interfacing logic an the arm lift motor rivers. Only some interfacing logic has been evelope as both the microcontroller basic eployment (R232 interface, quartz oscillator, reset circuitry) an the motor rivers have been foun on the market. The microcontroller must control the platter motor (start/stop) an the arm lifts (up/own). To report the turntable status to the transcription station it must also be able to rea the output of all the sensors an to measure the rotation spee of the platter. urthermore, as the start/stop signal accepte by the Technics motor control logic is of impulsive nature, a feeback signal must be erive to inform the mocrocontroller on the actual status of the motor. Examining the L-1210 schematics, the only reliable signal foun, avoiing to interfere with the normal motor operation, is a signal that goes high when the

14 14 PRETO IT D4.3 platter brakes are activate right after a stop comman is issue. This signal has been interface to an interrupt pin of the micro, so that its corresponing event is capture regarless of the mocro activity. The platter rotation spee can be measure by connecting to the Capture moule, built in the microcontroller, a signal generate by an inuctor triggere by a set of magnets positione uner the platter boy. The Capture moule is a peripheral able to continuously measure the time elapse between successive up fronts of a signal. The measurement has a irect relation with the rotation perio of the platter, as efine in the Computer Interface section (see for etails igure 8, igure 9 an igure 10). igure 8. Turntable control logic schematics. igure 9. Arm lift motor position sensors logic.

15 PRETO IT D igure 10. Microcontroller boar. igure 11. The control harware. This box is screwe uner the frame bottom. Microcontroller firmware The flow of the microcontroller firmware is escribe in the block iagrams below. At reset, after the initialisation of local variables an peripherals, the program loops in an ile cycle

16 16 PRETO IT D4.3 waiting for commans from the controlling station (igure 12). Upon reception of a byte on the serial interface, the value is analyse to see if it is a vali comman, if not it is rejecte an a NACK reply is sent back on the serial interface, otherwise the comman is execute. If the comman is a status request, a status byte is assemble by reaing the value of arm lift an position sensors an immeiately sent back on the interface; analogously, if the comman is a request of platter rotation spee, the Capture Register is rea an the low or high byte copie on the interface accoring to the request: it is up to the controlling station to reassemble the two bytes an convert the value in rouns per minute (RPM). If the comman is a start or a stop, then a ACK reply is immeiately sent back an the comman execution is then starte (igure 13). The comman execution consists of riving up or own the corresponing arm an of starting or stopping the platter motor, taking also into account the status of the other arm. Complete the comman execution, the program returns to its ile loop waiting for a new comman to execute (igure 14). Initialisation New comman? no yes Parse comman igure 12. Main comman loop. Parse comman yes Is a TARTn? en ACK Execute TARTn no Is a TOPn? yes en ACK Execute TOPn no yes Is a TATU? en TATU no Return en NACK no Is a PEED? yes en PEED Return igure 13. Comman parsing.

17 PRETO IT D Execute TARTn Execute TOPn Motor stoppe? no Raise arm n yes tart motor All arms up? no yes Lower arm n top motor Return Return igure 14. tart an top commans execution. 78 RPM spee implementation The Technics L-1210 spee is regulate by a AN6680 chip, accoring to a reference frequency generate by a crystal oscillator. Moifying this frequency, it is possible to implement any wante rotation spee. The moifications that nee to be applie to the L control are shown in igure 8. The switch selects the reference frequency source: on boar for 33/45 RPM or coming from the ae oscillator for 78 RPM. When the latter is selecte, the two switches on the eck for selecting 33 or 45 RPM are isable. This conition is flagge by switching both the corresponing les on. igure 15. Harware moification for 78 RPM implementation Turntable et up an Tuning The turntable set up epens on the type of recors that must be processe by the transcription chain. The following configurations are possible: ingle ME 309 tonearm with either or for 78 RPM recors. Recors of size greater than 12'' will not be playable.

18 18 PRETO IT D4.3 ingle ME 312 tonearm with cartrige an stylus either for 33/45 RMP recors or for 78 RPM recors. Recors of up to 16'' can be playe. Two ME 312 tonearms, one with cartrige an stylus for 33/45 RMP recors an one for 78 RPM recors. Recors of up to 16'' can be playe. Two ME 312 tonearms with ifferent styli for 78 RPM recors. Recors of up to 16'' can be playe. igure 16. The assemble turntable. The tuning of the unit requires proper trimming of the tonearm(s), following the instruction given in the ME 300 eries manual, an careful levelling of the eck that must be laying perfectly horizontal to avoi the generation of any unwante skating forces on the pick up. The levelling can be one, with the help of a spirit-level lai on the platter, by slightly turning the turntable feet. The trimming of the tonearms is more time consuming an must be one by an expert technician with the help of the arm manual. inally, the arm lift must be positione in such a way that it lifts the arm at least 5 mm over the recor when in rest position an that it oesn't interfere with it when in playing position. The regulation can be one either by sliing appropriately the sensors rees along the shaft of the motors, or by easing the screw that hols the ro of the lift on the motor shaft an moving it up or own as neee urther Design Optimisation Ieas While planning the builing of small series of the turntable, it can be consiere the opportunity of looking for further project optimisations, both in the irection of improving the turntable characteristics an reucing the implementation costs. Here are some preliminary ieas: Integration of the turntable eck in the frame bottom, that is eriving the riving motor, electronics an platter housing irectly in the aluminium block use for the frame base. This woul further increase the rigiity of the whole an the rejection to microvibrations. election of smaller arm lift motors. The motor use are actually conceive for higher workloas an are over imensione for this use. maller motors woul reuce costs an vibrations. Integration of the control harware. or the prototype implementation purposes, evaluation boars provie by manufacturers of eveloping tools have been use. The esign of a custom boar woul reuce costs an size of the electronics.

19 PRETO IT D A/D conversion technology selection 4.1. Key Link escription All the moern implementations of auio Analogue to Digital (A/D) converters use evices base on the igma-delta moulation technique that solves several problems generally foun on traitional parallel converters. igma-delta converters are in fact intrinsically linear over all the usable ynamic range an several evices on the market show that with this technology true 20 bit resolution can be obtaine, giving a Total Harmonic Distortion + Noise (THD+N) figure of. Recently, a new approach to A/D conversion has been propose, base on the use of a parallel of up to 4 igma-delta converters, appropriately scale to cover istinct ynamic ranges. The output of the single converters is then combine to obtain the igital representation of the analogue signal. The philosophy of this approach is that of extening the ynamic range of the converter up to 150, while maintaining the THD+N constant at about in a range of 50 of the input signal. Having iverging measurement results, namely higher ynamic range an lower THD+N, it is not trivial to instrumentally etermine if this approach brings any benefit to the quality of the conversion compare to that employing a single igma-delta converter, but further investigations, incluing listening tests are neee to unerstan how the perceptibility of the quantisation noise/istortion prouce by this equipment compares to that of state of the art single converter equipment. The work of this key link has been organize into three phases: election of state of the art equipment to test Objective measurements ubjective evaluations In rawing the conclusions of this analysis special attention will be given to istinguish, when possible, between ifferences ue to the specific implementation of the teste equipment an ifferences peculiar to the ifferent conversion approach, as only the latter are relevant for the key link Equipment selection The market offer of A/D igma Delta conversion units is very ample, an several manufacturers prouce state of the art equipment. We ecie to inclue in the test the Apogee PX, because it is a popular apparatus, well known in the broacast environment not only for its conversion quality but also for its flexibility. It offers, in fact, 2 channels at 24 bits sample at up to 96 k, several igital output interface options, quality monitor D/A conversion an a selection of ithering an synching options. or the etails, please refer to table 1. The qua igma Delta configuration is patente an offere only by tagetec, a German company well known for igital routers an mixers. Recently, they ecie to prouce a stan alone conversion unit, name TrueMatch RMC, erive from the A/D evelope to equip their igital mixers. esie the stanar line level inputs, this unit can be equippe with microphone inputs, thus avoiing the nee for external preamplifiers. It can be configure with 8, 16 or 24 inputs at 28 bits. It must be note, however, that the lowest 4 bits of each sample can be obtaine only when interfacing the unit with another tagetec prouct via a custom interface, as the stanar auio igital interfaces hanle only up to 24 bits. Therefore, all the tests that we have performe are one using 24 bit output. Refer to table 2 for the etails of the TrueMatch characteristics.

20 20 PRETO IT D4.3 Apogee PX features Two channels 24-bit A/D an D/A conversion with 117 ynamic range ampling requency at 44.1, 48, 88.2 an 96 k uilt in interfaces: AE, ADAT, TDI, /PDI (coax an optical) Apogee it-plitting allows 16-bit recorers to be use to store 24-bit an even 24/96 k signals Operating moes: Confience Monitor, Digital Copy, Analog Monitor. Apogee oft Limit Apogee's UV22 ithers 24-bit signals to 20 or 16 bits 2-channel LED bar metering User-efinable over etection (1-4 igital ) AE/EU istribution amplifier capability alance or unbalance analog I/O Gol-plate XLR jacks for analog an AE/EU I/O RCA jacks for /PDI in an out Two proprietary Apogee low-jitter master clocks Optional vieo sync moule tagetec TrueMatch RMC features 8/12/16/24 microphone inputs, 28-bit resolution 153 (RM) converter ynamics balance transformer-insulate XLR inputs phantom power supplie subsonic filter with ajustable cutoff frequency phase inversion ajustable igital gain output format: AE/EU, /PDI noise shaping available, re-quantization to 20 or 16 bits with ithering worclock input for external synchronization (NC connector) worclock output (NC connector), suitable as master clock balance transformer-insulate outputs serial computer interface (R232; 25-pole sub-d connector) an U connectors (epening on version): XLR male, NC, D-ub, Toslink/DNP, Cinch/RCA 4.3. Objective measurements To verify the technical specifications provie by the manufacturers of the two equipment an to be able to compare them on an objective basis, an extensive set of measurements have been performe. The measurements have been one using the Auio Precision ystem 2, a state of the art auio measurement equipment, wiely use by auio experts. oth the AD converters have been connecte to the ystem 2 using their AE/EU igital output. As the AE/EU interface is capable of carrying up to 24 bits per auio sample, the 28 bit capability of the TrueMatch has not been assesse. On the other han, as alreay state, 28 bit connection is available only by using non stanar interfaces an is therefore harly usable in a mixe venor environment. The complete set of measurements is reprouce below with a short interpretation of the results.

21 PRETO IT D T 1k This measure consists of computing the ast ourier Transform (T) of a pure 1 k sine tone converte by the unit. With this measurement, by varying the amplitue of the test signal, it is possible to estimate the maximum input level accepte by the apparatus, as the signal spectrum is heavily affecte by out of range signals. urthermore, it shows the istribution of the backgroun noise an the presence of istortion components. The graphics below show that the PX100 can accept signals at up to 19.9 U, while at 20 U the istortion is clearly unacceptable. The TrueMatch instea can go up to 22 U. This values give the igital full scale level, generally inicate with 0, of the two equipment. As there is a mismatch of 2 between the two units, all the measures will be execute correcting the input levels accoringly. The noise floor is locate at for all the measure conitions. Unfortunately, this is ue to the limit in resolution of the ystem 2. It is then to be expecte that the real noise floor of the equipment be at a lower value. Apogee PX100 TAGETEC TrueMatch RMC k 5k 7.5k 10k 12.5k 15k 17.5k 20k 22.5k k 5k 7.5k 10k 12.5k 15k 17.5k 20k 22.5k 1k. Input level +18 u. 1k. Input level +18 u. Apogee PX100 Apogee PX k 5k 7.5k 10k 12.5k 15k 17.5k 20k 22.5k k 5k 7.5k 10k 12.5k 15k 17.5k 20k 22.5k 1k. Input level u. 1k. input level +20 u. TAGETEC TrueMatch RMC k 5k 7.5k 10k 12.5k 15k 17.5k 20k 22.5k 1k. Input level +22 u.

22 22 PRETO IT D4.3 requency response The frequency response has been calculate at full scale an 16 below, to verify if the input level affects the measure. The results are extremely goo for both the units, as is generally the case with toay state of the art converters. Apogee PX100 TAGETEC TrueMatch RMC k 2k 5k 10k 20k requency response. Input level +20 u k 2k 5k 10k 20k requency response. Input level +22 u. Channel separation (crosstalk) The channel separation is compute by measuring the level at the output of a channel of the stereo pair while a sine test signal is presente at the input of the other. The measure has been performe at two amplitues: 1 an 16 below full scale. In this case the tagetec unit outperforms the Apogee, even if the values are very goo in both cases. -70 Apogee PX TATEGEC TrueMatch RMC k 2k 5k 10k 20k k 2k 5k 10k 20k Channel separation. Input level +4 u. Channel separation. Input level +6 u. Apogee PX100 TATEGEC TrueMatch RMC k 2k 5k 10k 20k k 2k 5k 10k 20k Channel separation. Input level +19 u. Channel separation. Input level +21 u. THD+N Total Harmonic Distortion + Noise (THD+N) is one of the most important an selective measures. It express the effective resolution power of the AD converter as function of the input level. Unfortunately, it cannot iscriminate between noise (wie spectrum) an harmonic istortion (frequencies locate a multiples of the input signal), therefore it is not possible to etermine from this measure the relate perceptual effect. Use T graphs to raw a qualitative estimation of the istribution of noise an istortion. To obtain a realistic

23 PRETO IT D estimation of the perception of the istortion an noise components, a weighting mask can be applie to the signal erive from the observation of sensibility threshols of the human ear. Generally, plotting the THD+N value function of the amplitue of the input signal, the shape of this curves is flat with some rise when the level approaches the maximum permitte. The PX100 follows this shape, while the TrueMatch has a quite peculiar behaviour, ue to its non conventional architecture: for input levels above 0 U its istortion is consistently 4 higher than that of the PX100, but for lower input levels it improves. Keeping constant the amplitue an varying the frequency, we can still observe a flat curve for amplitues several uner the maximum level, while approaching full scale both the units show some irregularity an pronounce ifferences between the left an right channel. Apogee PX100 TAGETEC TrueMatch RMC r r 1 k. 0 r = +20 u. 1 k. 0 r = +22 u. Apogee PX100 TAGETEC TrueMatch RMC k 2k 5k 10k 20k k 2k 5k 10k 20k THD+N. Input level +10 u. THD+N. Input level +10 u. Intermoulation Distortion (IMD) Intermoulation istortion gives a measure of istortion proucts not harmonically relate to the pure signal. This is important since these artifacts make music soun harsh an unpleasant. This measure can be consiere complementary to THD as here the frequencies sought are those prouce by the combine effect of two or more tones, that are therefore multiples of the ifference of the frequency of each pair of tones. There are several methos to measure the IMD, incluing MPTE, CCI an DIM. The first two are base on the combination of two sine tones, while the latter uses a combination of a sine tone with a square one. All of them have been compute, but for brevity we report here only the MPTE, which is slightly easier to rea than the others. The results are however substantially confirme by the other methos. MPTE IMD uses a test signal mae of a 60 tone combine with a 7 k tone in the ratio 4:1. The analysis is one by observing the presence an amount of intermoulation terms aroun the 7 k frequency, represente by frequencies locate at 7 k ± n*60. The T of the output signal has been plotte at two input levels: 1 an 16 below full scale. rom the resulting graphs it can be observe that the TrueMatch presents higher istortion than the PX100, especially at -16. ut looking at the next graph where the result of the test has been plotte sweeping the input level from 0 to -50, the conclusions are contraictory. The two units present similar behaviours in the range between -2 an -15 but then the TrueMatch curve becomes very irregular an

24 24 PRETO IT D4.3 significantly worse than that of the PX100 until -25, where the situation reverses while going towars the lower levels. Its is then very ifficult to preict what coul be the impact on perception of this unusual characteristic. Apogee PX100 TAGETEC TrueMatch RMC r MPTE ynamic intermoulation. 0 r = +20 u r MPTE ynamic intermoulation. 0 r = +22 u. Apogee PX100 TAGETEC TrueMatch RMC k 6.6k 6.7k 6.8k 6.9k 7k 7.1k 7.2k 7.3k 7.4k 7.5k k 6.6k 6.7k 6.8k 6.9k 7k 7.1k 7.2k 7.3k 7.4k 7.5k T MPTE Intermoulation. Input level +19 u. T MPTE Intermoulation. Input level +21 u. Apogee PX100 TAGETEC TrueMatch RMC k 6.6k 6.7k 6.8k 6.9k 7k 7.1k 7.2k 7.3k 7.4k 7.5k k 6.6k 6.7k 6.8k 6.9k 7k 7.1k 7.2k 7.3k 7.4k 7.5k T MPTE Intermoulation. Input level +4 u. T MPTE Intermoulation. Input level +6 u. Linearity Linearity is a measure of how much the converter steps iffer from the theoretical values. With moern igma-delta moulation base converters, this measure is useful to estimate the practical ynamic range of the equipment, because the esign metho ensures the linearity in the intermeiate steps. Two curves are rawn in the graphics: the iagonal blue ones represent the converte value of the input level, while the re an green ones show the ifference between the converter output an its theoretical value expresse in for the two channels of the stereo pair. The PX100 has an effective ynamic range of about 120, corresponing to 20 bits, similar to its istortion level shown by the THD+N figure (consiering its best channel), but the TrueMatch has a ynamic range of more than 140, corresponing to almost 24 bits, much higher than its THD+N value, that coul not be achieve even in theory, ue to the thermical noise of analog components.

25 PRETO IT D This measure, combine with the THD+N, show that the TrueMatch behaviour is substantially inepenent from the input level of the signal, while the PX100 presents a peestal at about -90 u that limits the converter resolution towars the low levels. Apogee PX100 TAGETEC TrueMatch RMC r r Linearity curve. 0 r = +20. u. Linearity curve. 0 r = +22u ubjective tests Objective It is well known that objective measurements can give only partial information about the overall quality of soun. This is mainly ue to the fact that objective measurements cannot account for the plethora of non linear couplings that occur when complex souns (in the meaning of not pure sine tones) are processe by an electro-acoustic chain an by the human ear. Therefore, listening is the only means that we have to rate the quality of equipment when the objective measurements give contraictory information, namely in our case: better ynamic range an worse istortion in some part of the auio ynamic range. The questions that we try to answer with this subjective experiment are: a) Is the ifferent istortion characteristic (THD+N an IMD) of the two units a iscriminator in the transcription of archive materials application? b) Is the extene ynamic range of the TrueMatch bringing a valuable plus to the conversion in the archive application? c) Is one of the two units globally better performing than the other? Test methoology The tests have been performe using the expert panel metho, wiely aopte by the broacast community to evaluate the quality of prouction equipment. It consists of comparative observation of source an processe material in varying listening conitions by a panel of expert listeners. The listeners are free to talk to each other an to exchange opinions. This metho is less rigorous than the classical ouble blin test, use by both auio an vieo communities to qualify coing systems, but, when the quality of the signals to be evaluate is very high, it generally prouces vali results at a much lower cost. or these tests, we gathere a panel of 5 expert listeners, 3 working in the auio group of the RAI Research Centre an the others auio technicians of the RAI Prouction Centre of Turin. Test conitions The two A/D converters have been compare in the following conitions: 1. live recoring of a performance of the RAI symphonic orchestra 2. transcription of various samples of archive materials (tapes an vinyl) 3. transcription of archive materials at levels significantly lower than the nominal

26 26 PRETO IT D4.3 Conition 1 has been inclue to evaluate the igitisation fielity of the two unit for high quality, high ynamic range music. It consiste of a stereo recoring of a rehearsal session of the RAI orchestra in the Turin Giovanni Agnelli auitorium, performing travinskij's "Concert in D for violin an orchestra". The soun has been recore simultaneously in three versions, two igital at 24 bit/sample, 48 k, capture by the two units uner test, an an analog one recore with a 1/4'' tuer A812 on a high quality tape at 15 inch/s spee, use uring the listening as a reference. The set up of the recoring is shown in igure 17. Note that, even if the TrueMatch converter can be connecte irectly to the microphones, a preamplifier has been use instea in orer to match as close as possible the recoring conition of the PX100. tereo Microphone MIDA XL42 Passive plitter TAGETEC TrueMatch RMC TUDER * Apogee PX100 A 812 AE EU ounscape Mixtreme car igure 17. Live recoring set up. * Tape spee 15 inch/s Conition 2 is the simulation of the normal working conition expecte in our application. It is significantly ifferent from the former case because here the ynamic range of the signal is much lower ue to the intrinsic limitations of the legacy analog recoring meia. everal types of supports an content have been use, incluing classical an pop 33 RPM vinyl from various perios an magnetic tape 1/4'' recorings of music, see the table below. Conition 3 has been ae to stress the equipment an verify the usability of the extene ynamic range of the TrueMatch. This conition can be meaningful in simulating mass transcription chains when some material can have been recore on the original meia at a non stanar level an the realignment of the equipment at each single transcription is unpractical. Here the source consiste of the same material use for conition 2 attenuate by 30 before the A/D input. The igital signal has then been realigne at nominal level before playback, in orer to avoi ifferences in perception ue to the ifferent listening level. Author Title erial number ARRY WHITE LET THE MUIC PLAY A DIRE TRAIT LOVE OVER GOLD DEMI ROUO HAPPY TO E A AUTO PAPETTI AX 20a RACCOLTA Ms AL LOREDANA ERTE JAZ C PATCWORK THE PHANTOM LP 704 AIO AOR PAPE ATAN HLP 200 CHOPIN NOTTURNI ZKY ONEY M. NIGHTLIGHT TO DAI GAZEO GAZEO R 56050